Data reconstruction in 2017/2018

In 2017/2018 the data reconstruction is much more complicated than before, mainly due to 13 time samples of the pulse sampling in HG/LG of the Skiroc2cms and the high noise observed in the beam period. The reconstruction step and over view is written based on the 2018 June framework. At the time of writing, several changes have been made in the reconstruction steps in the latest framework developed for 2018 October test beam.

However, the idea of the reconstruction steps remain the same.

The first step after decoding is to connect the data in the memory map to the sensor cells (DetectorID). All the connected channel are assigned a DetectorID, while the un-connected channel are assigned invalid DetectorID. In the pulse reconstruction step, the cells with invalid DetectorID are dropped.

Among the 3 circuits used in the energy measurement of the Skiroc2cms (HG/LG/TOT), the first two are linear circuits and have 13 samples to the pulse. The TOT circuit will only take part in the last step of the reconstruction, which is the gain choosing step.

The switch capacitor array (SCA) is the fundamental hardware unit in the Skiroc2cms HG/LG measurement. They can be consider as a clock with 13 numbers, which will be charged by a clock-wise running conductive pointer. The SCA unit can receive the pulse from the analog output and sliced the pulse into 25 ns interval and store the charge in the end. When the trigger signal arrives, the charge transmitting process is stopped and the digitization process will starts. Based on the stopping point of the pointer, the time sample (TS) of the event is assigned. See Figure 7.4. The order of the time sample is universal in the same module. If the trigger signal is not received, the charge of the last SCA will be dumped out and reset to achieve the continuous data taking. There is no

Figure 7.4: Pulse,TS and SCA

The concept of the TS and SCA is shown in the left, the C1, C2 ... are the capacitors in the fixed place while the black numbers are the TS which is only assigned when the trigger signal arrive. The right plot, shows the simple concept of the sampling.

buffer design to store the data more than 13 time samples (275 ns). This is a important constrain when designing the path of the trigger system. If the latency for trigger signal is not short enough, the information of the physical pulse is lost. The latency of the

2017/2018 setup is hence compressed to ∼ 175 ns. Thanks to the stability of the pulse shaper, the pulse peak sits very stable between TS 2 and TS 3 (start from 0).

The pedestal value is obtained from the SCA unit in a run-by-run basis. In a single SCA, only the TS 0 is considered in contributing the ADC distribution in order to remove the bias of the pulse. ^† From this ADC distribution the median is extracted to represent the given DC shift of the SCA. Each event is then subtracted by the corresponding SCA and rearranged to the order of TS.

Common mode evaluation is done on the cell type (The full, half, mousebite, etc) at the TS-by-TS basis. After selecting the cells with signal, the 1D distribution of single TS on the same CHIP is filled. Again the median of the distribution is chosen to represent the CM of such TS in the event.

So far the reconstruction can be expressed as:

sig(CH, SCA) − P ED(run, SCA) = sig(CH, T S) (7.3) sig(CH, T S) − CM (ev, T S, type) = Rawsig(CH, T S) (7.4) After the pedestal and CM removal, the data of the pulse is stored into a ”RAWhit”

format. The pulse is then fitted with a shape describe in 7.5 after the pre-selection that the peaking TS should lies in TS 2 or TS 3 and the value should be greater than 20 HG ADC.

The constants n,trise,τ and normalization factor is obtained from lab data. The fitting amplitude is the chosen number to represent the HG and LG measurement.

The fitting result of the HG and LG circuits are 2 single numbers, these 2 numbers as well as the TOT measurement are 3 results in a single cell. In the gain calibration step, the most reasonable value will be chosen to represent the energy of the cell in that event.

The gain calibration is performed based on studying the HG-LG and LG-TOT corre-lation. Since there is no way to obtain the real signal amplitude in the data, the study is a indirect method to find the overlap region between different gains.

By analyzing the correlation channel to channel, the transition point as well as the gain factor between HG/LG and LG/TOT is provided. Moreover, the TOT offset and is calculated. From the efforts in the gain study, converting all gains to a relative HG measurement is realized.

The last step will be the MIP calibration, which convert the HG ADC to the unit of MIP. The MIP calibration is again achieved by the muon runs. The good MIP events is selected based on the agreement of the results from DWCs tracking algorithm and the HGCAL hit position. The behavior of the MIP signal is analyzed channel by channel and shows the result in a lookup table.

†The TS 0 is considered to remain unaffected since the physical pulse has not arrived.

By the cooperation of the MIP calibration and gain calibration, the signal on the cell is finally converted into MIP units.

Sig(CH, convertedHG)

HG_{M IP}(CH) = Sig(M IP s) (7.6)

The data reconstruction of 2017/2018 is a story of optimizing and trouble shooting for the new developed electronics. The great success have been achieved since the first beam test in May, 2017.

The milestones of the reconstruction are listed below:

• The interposer is firstly applied as a bridge between the high/low voltage supply and the module, which improved the ADC distribution in 2017 june.

• The reason of the high total noise in the H2 beam line is pointed to the grounding of the bias voltage of the module. In 2018 March, method is provided to improve the grounding and applied to later test beams. The total noise drop is significant (From 200 ADC in HG to 20 ADC.)

• Thanks to the drop of the total noise, in 2018 June the MIP peak is shown in the LG.